Prediction of inhibitory activities of small molecules against Pantothenate synthetase from Mycobacterium...

The new and novel drug molecules are of prime importance against the deadly Mycobacterium tuberculosis owing to its high resistance. The discovery of new drug molecules is cost, time, and efforts intensive in chemical research. Computational approaches, such as virtual screening and Machine Learning represent an effective alternate to predict the active compounds with appreciable accuracy. In this work, we used the true active and in-active drug candidates to train the machine learned models against one of the potent drug targets from Mycobacterium tuberculosis i.e. Pantothenate synthetase (PS). We computed 1444 descriptors from the studied molecules. Initially, twenty descriptors were shortlisted based on their significant Pearson's correlation with the -logIC₅₀ values. Different combinations of descriptors were used to optimize the number of descriptors. Further to that different Machine Learned models were applied to develop a trained model of active molecules with a reasonable accuracy. The best performed model in terms of prediction of the activity data is proposed as a model of choice to perform the data screening experiments. The current study will help to potentiate the drug discovery process against Mycobacterium tuberculosis (Mtb).